PinnedSundar NarayananThe Need for Explainability and Simulation in Preference OptimizationThe world is multilingual and technology is recognizing the need to be multilingual. Working on multilingual data or models with language…Jul 10Jul 10
Sundar NarayananUnderstanding limitations, evaluating Risks, and Addressing Challenges in Retrieval-Augmented…Despite the potential benefits of retrieval augmented generation (RAG) in terms of generating accurate and contextually relevant responses…Mar 28Mar 28
Sundar NarayananBias Risk Assessment — A systematic approach — Part 2/3The primary step to assess bias risk would require gathering preliminary information regarding the AI System. To gather such preliminary…Nov 28, 2022Nov 28, 2022
Sundar NarayananBias Risk Assessment — A systematic approach — Part 2/3Bias risk assessment shall consider the types of bias, types of data and types of models prior to contextually assessing them for…Nov 28, 2022Nov 28, 2022
Sundar NarayananBias Risk Assessment — A systematic approach — Part 1/2Bias is contributed by data, model, outcome, interface, environment, and human actions in which such AI/ML system is used. It could exist…Nov 28, 2022Nov 28, 2022
Sundar NarayananEssential terms and conditions for public procurement of Artificial Intelligence related products…The government published a National Strategy for Artificial Intelligence in June 2018 and launched a National AI portal (here) containing a…Nov 25, 2022Nov 25, 2022
Sundar NarayananI know how you swipe on your device!!!Dealing with the risks of behavior biometricsJun 29, 2022Jun 29, 2022
Sundar NarayananHuman-in-the-loop or on-the-loop is not a silver bullet. Evaluate their effectiveness!!!Framework for evaluating effectiveness of Human-in-the-loop or Human-on-the-loopJan 4, 20221Jan 4, 20221
Sundar NarayananinTowards Data ScienceGitHub and Pre-Trained Models: A Keyhole ViewGitHub contains pre-trained models that may contain inherent biases, and the absence of model scorecards will contribute to amplifying the…Jul 12, 2021Jul 12, 2021
Sundar NarayananinTowards Data ScienceA cautionary tale: multi-stakeholder feedback in AI EthicsMulti-stakeholder feedback has its inherent flaws, hence, its best to treat them as ‘one of the means’ than an ‘as an end’ to establishing…Jun 30, 2021Jun 30, 2021